DNS HOWTO <author>Nicolai Langfeldt <tt/janl@math.uio.no/ <date>v1.4.2, 1 January 1998 <abstract> HOWTO become a totally small time DNS admin. </abstract>  <toc> <sect>Preamble Keywords: DNS, bind, named, dialup, ppp, slip, Internet, domain, name, hosts, resolving <sect1>Legal stuff (C)opyright 1995 Nicolai Langfeldt. Do not modify without amending copyright, distribute freely but retain copyright message. <sect1>Credits and request for help. I want to thank Arnt Gulbrandsen who read the drafts to this work countless times and provided many useful suggestions. I also want to thank the people that have e-mailed suggestions, and thank you notes. This will never be a finished document, please send me mail about your problems and successes, it can make this a better HOWTO. So please send money, comments and/or questions to janl@math.uio.no. If you send e-mail and want an answer please show the simple courtecy of <em/making sure/ that the return address is correct and working. Also, <bf/please/ read the <ref id="qanda" name="FAQ"> section before mailing me. If you want to translate this HOWTO please notify me so I can keep track of what languages I have been published in :-). <sect1>Dedication This HOWTO is dedicated to Anne Line Norheim Langfeldt. Though she will probably never read it since she's not that kind of girl. <sect>Introduction.<label id="intro"> <bf/What this is and isn't./ For starters, DNS is is the Domain Name System. The rules that name machines and software that maps those names to IP numbers. This HOWTO documents how to define such mappings using a Linux system. A mapping i simply a association between two things, in this case a machine name, like ftp.linux.org, and the machines IP number, 199.249.150.4. DNS is, to the uninitiated (you ;-), one of the more opaque areas of network administration. This HOWTO will try to make a few things clearer. It describes how to set up a <em/simple/ DNS name server. Starting with a caching only server and going on to setting up a primary DNS server for a domain. For more complex setups you can check the <ref id="qanda" name="FAQ"> section of this document. If it's not described there you will need to <em/read/ the Real Documentation. I'll get back to what this Real Documentation consists of in <ref id="bigger" name="the last chapter">. Before you start on this you should configure your machine so that you can telnet in and out of it, and make successfully make all kinds of connections to the net, and you should especially be able to do <tt/telnet 127.0.0.1/ and get your own machine (test it now!). You also need a good <tt>/etc/host.conf</tt> (or <tt>/etc/nsswitch.conf</tt>), <tt>/etc/resolv.conf</tt> and <tt>/etc/hosts</tt> files as a starting point, since I will not explain their function here. If you don't already have all this set up and working the networking/NET-2 HOWTO explains how to set it up. Read it. If you're using SLIP or PPP you need that working. Read the PPP HOWTO if it's not. When I say `your machine' I mean the machine you are trying to set up DNS on. Not any other machine you might have that's involved in your networking effort. I assume you're not behind any kind of firewall that blocks name queries. If you are you will need a special configuration, see the section on <ref id="qanda" name="FAQ">. Name serving on Unix is done by a program called <tt/named/. This is a part of the bind package which is coordinated by Paul Vixie for The Internet Software Consortium. <tt/Named/ is included in most Linux distributions and is usually installed as <tt>/usr/sbin/named</tt>. If you have a named you can probably use it; if you don't have one you can get a binary off a Linux ftp site, or get the latest and greatest source from <htmlurl url="ftp://ftp.isc.org/isc/bind/src/cur/" name="ftp.isc.org:/isc/bind/src/cur/"> This howto is about bind version 4. If you install version 8 you're almost on your own. There is a section later to help you a bit though. DNS is a net-wide database. Take care about what you put into it. If you put junk into it, you, and others will get junk out of it. Keep your DNS tidy and consistent and you will get good service from it. Learn to use it, admin it, debug it and you will be another good admin keeping the net from falling to it's knees overloaded by mismanagement. In this document I state flatly a couple of things that are not completely true (they are at least half truths though). All in the interest of simplification. Things will (probably ;-) work if you believe what I say. <bf/Tip:/ Make backup copies of all the files I instruct you to change if you already have them, so if after going through this nothing works you can get it back to your old, working state. <sect>A caching only name server.<label id="caching"> <bf/A first stab at DNS config, very useful for dialup users./ A caching only name server will find the answer to name queries and remember the answer the next time you need it. First you need a file called <tt>/etc/named.boot</tt>. This is read when named starts. For now it should simply contain: <code> ; Boot file for caching only name server ; directory /var/named ; ; type domain source file or host cache . root.cache primary 0.0.127.in-addr.arpa pz/127.0.0 </code> <bf/VERY IMPORTANT:/ In some versions of this document the file contents listed here will have a couple of spaces or a tab before the first non blank character. These are not supposed to be in the file. <bf/Delete any leading space/ in the files you cut and paste from this HOWTO. The `<tt/directory/' line tells named where to look for files. All files named subsequently will be relative to this. <tt>/var/named</tt> is the right directory according to the <em/Linux File system Standard/. Thus <tt>pz</tt> is a directory under <tt>/var/named</tt>, i.e., <tt>/var/named/pz</tt>. The file named <tt>/var/named/root.cache</tt> is named in this. <tt>/var/named/root.cache</tt> should contain this: <code> . 518400 NS D.ROOT-SERVERS.NET. . 518400 NS E.ROOT-SERVERS.NET. . 518400 NS I.ROOT-SERVERS.NET. . 518400 NS F.ROOT-SERVERS.NET. . 518400 NS G.ROOT-SERVERS.NET. . 518400 NS A.ROOT-SERVERS.NET. . 518400 NS H.ROOT-SERVERS.NET. . 518400 NS B.ROOT-SERVERS.NET. . 518400 NS C.ROOT-SERVERS.NET. ; D.ROOT-SERVERS.NET. 3600000 A 128.8.10.90 E.ROOT-SERVERS.NET. 3600000 A 192.203.230.10 I.ROOT-SERVERS.NET. 3600000 A 192.36.148.17 F.ROOT-SERVERS.NET. 3600000 A 192.5.5.241 G.ROOT-SERVERS.NET. 3600000 A 192.112.36.4 A.ROOT-SERVERS.NET. 3600000 A 198.41.0.4 H.ROOT-SERVERS.NET. 3600000 A 128.63.2.53 B.ROOT-SERVERS.NET. 3600000 A 128.9.0.107 C.ROOT-SERVERS.NET. 3600000 A 192.33.4.12 </code> <bf/Remember what I said about leading spaces!/ The file describes the root name servers in the world. This changes over time and <em/must/ be maintained. See the <ref id="maint" name="maintenance section"> for how to keep it up to date. This file is described in the named man page, but it is, IMHO, best suited for people that already understand named. The next line in <tt/named.boot/ is the <tt/primary/ line. I will explain its use in a later chapter, for now just make this a file named <tt/127.0.0/ in the subdirectory <tt/pz/: <code> @ IN SOA ns.linux.bogus. hostmaster.linux.bogus. ( 1 ; Serial 28800 ; Refresh 7200 ; Retry 604800 ; Expire 86400) ; Minimum TTL NS ns.linux.bogus. 1 PTR localhost. </code> Next, you need a <tt>/etc/resolv.conf</tt> looking something like this: <code> search subdomain.your-domain.edu your-domain.edu nameserver 127.0.0.1 </code> The `<tt/search/' line specifies what domains should be searched for any host names you want to connect to. The `<tt/nameserver/' line specifies the address of your nameserver at, in this case your own machine since that is where your named runs. If you want to list several name servers put in one `<tt/nameserver/' line for each. (Note: Named never reads this file, the resolver that uses named does.) To illustrate what this file does: If a client tries to look up <tt>foo</tt>, <tt>foo.subdomain.your-domain.edu</tt> is tried first, then <tt>foo.your-fomain.edu</tt>, finally <tt>foo</tt>. If a client tries to look up <tt>sunsite.unc.edu</tt>, <tt>sunsite.unc.edu.subdomain.your-domain.edu</tt> is tried first (yes, it's silly, but that's the way it's gotta be) , then <tt>sunsite.unc.edu.your-domain.edu</tt>, and finally <tt>sunsite.unc.edu</tt>. You may not want to put in too many domains in the search line, it takes time to search them. The example assumes you belong in the domain <tt>subdomain.your-domain.edu</tt>, your machine then, is probably called <tt>your-machine.subdomain.your-domain.edu</tt>. The search line should not contain your TLD (Top Level Domain, `<tt/edu/' in this case). If you frequently need to connect to hosts in another domain you can add that domain to the search line like this: <code> search subdomain.your-domain.edu your-domain.edu other-domain.com </code> and so on. Obviously you need to put real domain names in instead. Please note the lack of periods at the end of the domain names. Next, depending on your libc version you either need to fix <tt>/etc/nsswitch.conf</tt> or <tt>/etc/host.conf</tt>. If you already have <tt/nsswitch.conf/ that's what we'll fix, if not, we'll fix <tt/host.conf/. <bf>/etc/nsswitch.conf</bf> This is a long file specifying where to get different kinds of data types, from what file or database. It usually contains helpful comments at the top, which you should consider reading, now. After that find the line starting with `<tt/hosts:/', it should read <code> hosts: files dns </code> If there is no line starting with `<tt/hosts:/' then put in the one above. It says that programs should first look in the <tt>/etc/hosts</tt> file, then check DNS according to <tt/resolv.conf/. <bf>/etc/host.conf</bf> It probably contains several lines, one should starting with <tt/order/ and it should look like this: <code> order hosts,bind </code> If there is no `<tt/order/' line you should stick one in. It tells the host name resolving routines to first look in <tt>/etc/hosts</tt>, then ask the name server (which you in <tt/resolv.conf/ said is at 127.0.0.1) These two latest files are documented in the resolv(8) man page (do `<tt/man 8 resolv/') in most Linux distributions. That man page is IMHO readable, and everyone, especially DNS admins, should read it. Do it now, if you say to yourself "I'll do it later" you'll never get around to it. <sect1>Starting named After all this it's time to start named. If you're using a dialup connection connect first. Type `<tt/ndc start/', and press return, no options. If that back-fires try `<tt>/usr/sbin/ndc start</tt>' instead. If that back-fires see the <ref id="qanda" name="FAQ"> section. Now you can test your setup. If you view your syslog message file (usually called <tt>/var/adm/messages</tt>, but another directory to look in is <tt>/var/log</tt> and another file to look in is <tt/syslog/) while starting named (do <tt>tail -f /var/adm/messages</tt>) you should see something like: <tscreen><verb> Jun 30 21:50:55 roke named[2258]: starting. named 4.9.4-REL Sun Jun 30 21:29:03 MET DST 1996 janl@roke.slip.ifi.uio.no:/var/tmp/bind/named Jun 30 21:50:55 roke named[2258]: cache zone "" loaded (serial 0) Jun 30 21:50:55 roke named[2258]: primary zone "0.0.127.in-addr.arpa" loaded (serial 1) </verb></tscreen> If there are any messages about errors then there is a mistake. Named will name the file it is in (one of named.boot and root.cache I hope :-) Kill named and go back and check the file. Now it's time to start nslookup to examine your handywork. <tscreen><verb> $ nslookup Default Server: localhost Address: 127.0.0.1 > </verb></tscreen> If that's what you get it's working. We hope. Anything else, go back and check everything. Each time you change the <tt/named.boot/ file you need to restart named using the <tt/ndc restart/ command. Now you can enter a query. Try looking up some machine close to you. <tt/pat.uio.no/ is close to me, at the University of Oslo: <tscreen><verb> > pat.uio.no Server: localhost Address: 127.0.0.1 Name: pat.uio.no Address: 129.240.2.50 </verb></tscreen> nslookup now asked your named to look for the machine <tt/pat.uio.no/. It then contacted one of the name server machines named in your <tt>root.cache</tt> file, and asked its way from there. It might take tiny while before you get the result as it searches all the domains you named in <tt>/etc/resolv.conf</tt>. If you try again you get this: <tscreen><verb> > pat.uio.no Server: localhost Address: 127.0.0.1 Non-authoritative answer: Name: pat.uio.no Address: 129.240.2.50 </verb></tscreen> Note the `<tt/Non-authoritative answer:/' line we got this time around. That means that named did not go out on the network to ask this time, it instead looked in it's cache and found it there. But the cached information <em/might/ be out of date (stale). So you are informed of this (very slight) danger by it saying `<tt/Non-authorative answer:/'. When <tt/nslookup/ says this the second time you ask for a host it's a sure sign it named caches the information and that it's working. You exit <tt/nslookup/ by giving the command `<tt/exit/'. If you're a dialup (ppp, slip) user please read the <ref id=auto name="section on dialup connections">, there is some advice there for you. Now you know how to set up a caching named. Take a beer, milk, or whatever you prefer to celebrate it. <sect>A <em/simple/ domain.<label id="simple"> <bf>How to set up your own domain.</bf> <sect1>But first some dry theory Before we <em/really/ start this section I'm going to serve you some theory on how DNS works. And you're going to read it because it's good for you. If you don't `wanna' you should at least skim it very quickly. Stop skimming when you get to what should go in your <tt/named.boot/ file. DNS is a hierarchical system. The top is written `<tt/./' and pronounced `root'. Under . there are a number of Top Level Domains (TLDs), the best known ones are ORG, COM, EDU and NET, but there are many more. When looking for a machine the query proceeds recursively into the hierarchy starting at the top. If you want to find out the address of <tt/prep.ai.mit.edu/ your name server has to find a name server that serves edu. It asks a <tt/./ server (it already knows the <tt/./ servers, that's what the <tt/root.cache/ file is for), the <tt/./ server gives a list of edu servers: <tscreen><verb> $ nslookup Default Server: localhost Address: 127.0.0.1 </verb></tscreen> Start asking a root server. <tscreen><verb> > server c.root-servers.net. Default Server: c.root-servers.net Address: 192.33.4.12 </verb></tscreen> Set the Query type to NS (name server records). <tscreen><verb> > set q=ns </verb></tscreen> Ask about edu. <tscreen><verb> > edu. </verb></tscreen> The trailing . here is significant, it tells the server we're asking that edu is right under . (this narrows the search somewhat). <tscreen><verb> edu nameserver = A.ROOT-SERVERS.NET edu nameserver = H.ROOT-SERVERS.NET edu nameserver = B.ROOT-SERVERS.NET edu nameserver = C.ROOT-SERVERS.NET edu nameserver = D.ROOT-SERVERS.NET edu nameserver = E.ROOT-SERVERS.NET edu nameserver = I.ROOT-SERVERS.NET edu nameserver = F.ROOT-SERVERS.NET edu nameserver = G.ROOT-SERVERS.NET A.ROOT-SERVERS.NET internet address = 198.41.0.4 H.ROOT-SERVERS.NET internet address = 128.63.2.53 B.ROOT-SERVERS.NET internet address = 128.9.0.107 C.ROOT-SERVERS.NET internet address = 192.33.4.12 D.ROOT-SERVERS.NET internet address = 128.8.10.90 E.ROOT-SERVERS.NET internet address = 192.203.230.10 I.ROOT-SERVERS.NET internet address = 192.36.148.17 F.ROOT-SERVERS.NET internet address = 192.5.5.241 G.ROOT-SERVERS.NET internet address = 192.112.36.4 </verb></tscreen> This tells us that <tt/*.root-servers.net/ serves <tt/edu./, so we can go on asking <tt/c/. Now we want to know who serves the next level of the domain name: <tt/mit.edu./: <tscreen><verb> > mit.edu. Server: c.root-servers.net Address: 192.33.4.12 Non-authoritative answer: mit.edu nameserver = STRAWB.mit.edu mit.edu nameserver = W20NS.mit.edu mit.edu nameserver = BITSY.mit.edu Authoritative answers can be found from: STRAWB.mit.edu internet address = 18.71.0.151 W20NS.mit.edu internet address = 18.70.0.160 BITSY.mit.edu internet address = 18.72.0.3 </verb></tscreen> <tt/steawb/, <tt/w20ns/ and <tt/bitsy/ serves <tt/mit/, select one and inquire about <tt/ai.mit.edu/: <tscreen><verb> > server W20NS.mit.edu. </verb></tscreen> Host names are not case sensitive, but I use my mouse to cut and paste so it gets copied as-is from the screen. <tscreen><verb> Server: W20NS.mit.edu Address: 18.70.0.160 > ai.mit.edu. Server: W20NS.mit.edu Address: 18.70.0.160 Non-authoritative answer: ai.mit.edu nameserver = WHEATIES.AI.MIT.EDU ai.mit.edu nameserver = ALPHA-BITS.AI.MIT.EDU ai.mit.edu nameserver = GRAPE-NUTS.AI.MIT.EDU ai.mit.edu nameserver = TRIX.AI.MIT.EDU ai.mit.edu nameserver = MUESLI.AI.MIT.EDU Authoritative answers can be found from: AI.MIT.EDU nameserver = WHEATIES.AI.MIT.EDU AI.MIT.EDU nameserver = ALPHA-BITS.AI.MIT.EDU AI.MIT.EDU nameserver = GRAPE-NUTS.AI.MIT.EDU AI.MIT.EDU nameserver = TRIX.AI.MIT.EDU AI.MIT.EDU nameserver = MUESLI.AI.MIT.EDU WHEATIES.AI.MIT.EDU internet address = 128.52.32.13 WHEATIES.AI.MIT.EDU internet address = 128.52.35.13 ALPHA-BITS.AI.MIT.EDU internet address = 128.52.32.5 ALPHA-BITS.AI.MIT.EDU internet address = 128.52.37.5 GRAPE-NUTS.AI.MIT.EDU internet address = 128.52.32.4 GRAPE-NUTS.AI.MIT.EDU internet address = 128.52.36.4 TRIX.AI.MIT.EDU internet address = 128.52.32.6 TRIX.AI.MIT.EDU internet address = 128.52.38.6 MUESLI.AI.MIT.EDU internet address = 128.52.32.7 MUESLI.AI.MIT.EDU internet address = 128.52.39.7 </verb></tscreen> So <tt/weaties.ai.mit.edu/ is a nameserver for <tt/ai.mit.edu/: <tscreen><verb> > server WHEATIES.AI.MIT.EDU. Default Server: WHEATIES.AI.MIT.EDU Addresses: 128.52.32.13, 128.52.35.13 </verb></tscreen> Now I change query type, we've found the name server so now we're going to ask about everything wheaties knows about <tt/prep.ai.mit.edu/. <tscreen><verb> > set q=any > prep.ai.mit.edu. Server: WHEATIES.AI.MIT.EDU Addresses: 128.52.32.13, 128.52.35.13 prep.ai.mit.edu CPU = dec/decstation-5000.25 OS = unix prep.ai.mit.edu inet address = 18.159.0.42, protocol = tcp #21 #23 #25 #79 prep.ai.mit.edu preference = 1, mail exchanger = life.ai.mit.edu prep.ai.mit.edu internet address = 18.159.0.42 ai.mit.edu nameserver = alpha-bits.ai.mit.edu ai.mit.edu nameserver = wheaties.ai.mit.edu ai.mit.edu nameserver = grape-nuts.ai.mit.edu ai.mit.edu nameserver = mini-wheats.ai.mit.edu ai.mit.edu nameserver = trix.ai.mit.edu ai.mit.edu nameserver = muesli.ai.mit.edu ai.mit.edu nameserver = count-chocula.ai.mit.edu ai.mit.edu nameserver = life.ai.mit.edu ai.mit.edu nameserver = mintaka.lcs.mit.edu life.ai.mit.edu internet address = 128.52.32.80 alpha-bits.ai.mit.edu internet address = 128.52.32.5 wheaties.ai.mit.edu internet address = 128.52.35.13 wheaties.ai.mit.edu internet address = 128.52.32.13 grape-nuts.ai.mit.edu internet address = 128.52.36.4 grape-nuts.ai.mit.edu internet address = 128.52.32.4 mini-wheats.ai.mit.edu internet address = 128.52.32.11 mini-wheats.ai.mit.edu internet address = 128.52.54.11 mintaka.lcs.mit.edu internet address = 18.26.0.36 </verb></tscreen> So starting at <tt/./ we found the successive name servers for the next level in the domain name. If you had used your own DNS server instead of using all those other servers, your named would of-course cache all the information it found while digging this out for you, and it would not have to ask again for a while. A much less talked about, but just as important domain is <tt/in-addr.arpa/. It too is nested like the `normal' domains. <tt/in-addr.arpa/ allows us to get the hosts name when we have it's address. A important thing here is to note that ip#s are written in reverse order in the in-addr.arpa domain. If you have the address of a machine: 192.128.52.43 named proceeds just like for the <tt/prep.ai.mit.edu/ example: find <tt/arpa./ servers. Find <tt/in-addr.arpa./ servers, find <tt/192.in-addr.arpa./ servers, find <tt/128.192.in-addr.arpa./ servers, <tt/find 52.128.192.in-addr.arpa./ servers. Find needed records for <tt/43.52.128.192.in-addr.arpa./ Clever huh? (Say `yes'.) The reversion of the numbers can be confusing the first 2 years. I have just told a lie. DNS does not work literally the way I just told you. But it's close enough. <sect1>Our own domain Now to define our own domain. We're going to make the domain <em/linux.bogus/ and define machines in it. I use a totally bogus domain name to make sure we disturb no-one Out There. We've already started this part with this line in <tt/named.boot/: <code> primary 0.0.127.in-addr.arpa pz/127.0.0 </code> Please note the lack of `<tt/./' at the end of the domain names in this file. The first line names the file <tt>pz/127.0.0</tt> as defining <tt/0.0.127.in-addr.arpa/. We've already set up this file, it reads: <code> @ IN SOA ns.linux.bogus. hostmaster.linux.bogus. ( 1 ; Serial 28800 ; Refresh 7200 ; Retry 604800 ; Expire 86400) ; Minimum TTL NS ns.linux.bogus. 1 PTR localhost. </code> Please note the `<tt/./' at the end of all the full domain names in this file, in contrast to the <tt/named.boot/ file above. Some people like to start each zone file with a <tt/$ORIGIN/ directive, but this is superfluous. The origin (where in the DNS hierarchy it belongs) of a zone file is specified in the `domain' column of the <tt/named.boot/ file, in this case it's <tt/0.0.127.in-addr.arpa/. This `zone file' contains 3 `resource records' (RRs): A SOA RR. A NS RR and a PTR RR. SOA is short for Start Of Authority. The `@' is a special notation meaning the origin, and since the `domain' column for this file says 0.0.127.in-addr.arpa the first line really means <tscreen><verb> 0.0.127.IN-ADDR.ARPA. IN SOA ... </verb></tscreen> NS is the Name Server RR, it tells DNS what machine is the name server of the domain <tt/0.0.127.in-addr.arpa/, it is <tt/ns.linux.bogus/. And finally the PTR record says that 1 (equals <tt/1.0.0.127.in-addr.arpa/, i.e. 127.0.0.1) is named <tt/localhost/. The SOA record is the preamble to <em/all/ zone files, and there should be exactly one in each zone file, the very first record. It describes the zone, where it comes from (a machine called <tt/linux.bogus/), who is responsible for its contents (<tt/hostmaster@linux.bogus/), what version of the zone file this is (serial: 1), and other things having to do with caching and secondary DNS servers. For the rest of the fields ,refresh, retry, expire and minimum use the numbers used in this HOWTO and you should be safe. Now restart your named (the command is <tt/ndc restart/) and use nslookup to examine what you've done: <tscreen><verb> $ nslookup Default Server: localhost Address: 127.0.0.1 > 127.0.0.1 Server: localhost Address: 127.0.0.1 Name: localhost Address: 127.0.0.1 </verb></tscreen> so it manages to get <tt/localhost/ from 127.0.0.1, good. Now for our main task, the <tt/linux.bogus/ domain, insert a new primary line in <tt/named.boot/: <code> primary linux.bogus pz/linux.bogus </code> Note the continued lack of ending `<tt/./' on the domain name in the <tt/named.boot/ file. In the linux.bogus zone file we'll put some totally bogus data: <code> ; ; Zone file for linux.bogus ; ; Mandatory minimum for a working domain ; @ IN SOA ns.linux.bogus. hostmaster.linux.bogus. ( 199511301 ; serial, todays date + todays serial # 28800 ; refresh, seconds 7200 ; retry, seconds 3600000 ; expire, seconds 86400 ) ; minimum, seconds NS ns.linux.bogus. NS ns.friend.bogus. MX 10 mail.linux.bogus ; Primary Mail Exchanger MX 20 mail.friend.bogus. ; Secondary Mail Exchanger localhost A 127.0.0.1 ns A 127.0.0.2 mail A 127.0.0.4 </code> Two things must be noted about the SOA record. ns.linux.bogus <em/must/ be a actual machine with a A record. It is not legal to have a CNAME record for he machine mentioned in the SOA record. It's name need not be `ns', it could be any legal host name. Next, hostmaster.linux.bogus should be read as hostmaster@linux.bogus, this should be a mail alias, or a mailbox, where the person(s) maintaining DNS should read mail frequently. Any mail regarding the domain will be sent to the address listed here. The name need not be `hostmaster', it can be any legal e-mail address, but the e-mail address `hostmaster' <em/is/ expected to work as well. There is one new RR type in this file, the MX, or Mail eXchanger RR. It tells mail systems where to send mail that is addressed to <tt/someone@linux.bogus/, namely too <tt/mail.linux.bogus/ or <tt/mail.friend.bogus/. The number before each machine name is that MX RRs priority. The RR with the lowest number (10) is the one mail should be sent to primarily. If that fails it can be sent to one with a higher number, a secondary mail handler, i.e. <tt/mail.friend.bogus/ which has priority 20 here. Restart named by running <tt/ndc restart/. Examine the results with nslookup: <tscreen><verb> $ nslookup > set q=any > linux.bogus Server: localhost Address: 127.0.0.1 linux.bogus origin = linux.bogus mail addr = hostmaster.linux.bogus serial = 199511301 refresh = 28800 (8 hours) retry = 7200 (2 hours) expire = 604800 (7 days) minimum ttl = 86400 (1 day) linux.bogus nameserver = ns.linux.bogus linux.bogus nameserver = ns.friend.bogus linux.bogus preference = 10, mail exchanger = mail.linux.bogus.linux.bogus linux.bogus preference = 20, mail exchanger = mail.friend.bogus linux.bogus nameserver = ns.linux.bogus linux.bogus nameserver = ns.friend.bogus ns.linux.bogus internet address = 127.0.0.2 mail.linux.bogus internet address = 127.0.0.4 </verb></tscreen> Upon careful examination you will discover a bug. The line <tscreen><verb> linux.bogus preference = 10, mail exchanger = mail.linux.bogus.linux.bogus </verb></tscreen> is all wrong. It should be <tscreen><verb> linux.bogus preference = 10, mail exchanger = mail.linux.bogus </verb></tscreen> I deliberately made a mistake so you could learn from it :-) Looking in the zone file we find that the line <tscreen><verb> @ MX 10 mail.linux.bogus ; Primary Mail Exchanger </verb></tscreen> is missing a period. Or has a 'linux.bogus' too many. If a machine name does not end in a period in a zone file the origin is added to it's end. So either <code> @ MX 10 mail.linux.bogus. ; Primary Mail Exchanger </code> or <code> @ MX 10 mail ; Primary Mail Exchanger </code> is correct. I prefer the latter form, it's less to type. In a zone file the domain should either be written out and ended with a `<tt/./' or it should not be included at all, in which case it defaults to the origin. Others have strong opinions going in the other direction. I must stress that in the named.boot file there should <em/not/ be `<tt/./'s after the domain names. You have no idea how many times a `<tt/./' too many or few have fouled up things and confused the h*ll out of people. So having made my point here is the new zone file, with some extra information in it as well: <code> ; ; Zone file for linux.bogus ; ; Mandatory minimum for a working domain ; @ IN SOA ns.linux.bogus. hostmaster.linux.bogus. ( 199511301 ; serial, todays date + todays serial # 28800 ; refresh, seconds 7200 ; retry, seconds 604800 ; expire, seconds 86400 ) ; minimum, seconds NS ns ; Inet Address of name server NS ns.friend.bogus. MX 10 mail ; Primary Mail Exchanger MX 20 mail.friend.bogus. ; Secondary Mail Exchanger localhost A 127.0.0.1 ns A 127.0.0.2 mail A 127.0.0.4 ; ; Extras ; @ TXT "Linux.Bogus, your DNS consultants" ns MX 10 mail MX 20 mail.friend.bogus. HINFO "Pentium" "Linux 1.2" TXT "RMS" richard CNAME ns www CNAME ns donald A 127.0.0.3 MX 10 mail MX 20 mail.friend.bogus. HINFO "i486" "Linux 1.2" TXT "DEK" mail MX 10 mail MX 20 mail.friend.bogus. HINFO "386sx" "Linux 1.0.9" ftp A 127.0.0.5 MX 10 mail MX 20 mail.friend.bogus. HINFO "P6" "Linux 1.3.59" </code> You might want to move the first three A records so that they're placed next to their respective other records, instead on top like that. There are a number of new RRs here: HINFO (Host INFOrmation) has two parts, it's a good habit to quote each. The first part is the hardware or CPU on the machine, and the second part the software or OS on the machine. ns has a Pentium CPU and runs Linux 1.2. The TXT record is a free text record that you can use for anything you like. CNAME (Canonical NAME) is a way to give each machine several names. So richard and www is a alias for ns. It's important to note that A MX, CNAME and SOA record should <em/never/ refer to a CNAME record, they should only refer to something with a A record, so it would wrong to have <code> foobar CNAME richard ; NO! </code> but correct to have <code> foobar CNAME ns ; Yes! </code> It's also important to note that a CNAME is not a legal host name for a e-mail address: <tt/webmaster@www.linux.bogus/ is an illegal e-mail address given the setup above. You can expect quite a few mail admins Out There to enforce this rule even if it works for you. The way to avoid this is to use A records (and perhaps some others too, like a MX record) instead: <code> www A 127.0.0.2 </code> Paul Vixie, the primary named wizard, recommends <em/not/ using CNAME. So consider not using it <em/very/ seriously. Load the new database by running <tt/ndc reload/, this causes named to read its files again. <tscreen><verb> $ nslookup Default Server: localhost Address: 127.0.0.1 > ls -d linux.bogus </verb></tscreen> This means that all records should be listed. <tscreen><verb> [localhost] linux.bogus. SOA ns.linux.bogus hostmaster.linux.bogus. (199511301 28800 7200 604800 86400) linux.bogus. NS ns.linux.bogus linux.bogus. NS ns.friend.bogus linux.bogus. MX 10 mail.linux.bogus linux.bogus. MX 20 mail.friend.bogus linux.bogus. TXT "Linux.Bogus, your DNS consultants" localhost A 127.0.0.1 mail A 127.0.0.4 mail MX 10 mail.linux.bogus mail MX 20 mail.friend.bogus mail HINFO 386sx Linux 1.0.9 donald A 127.0.0.3 donald MX 10 mail.linux.bogus donald MX 20 mail.friend.bogus donald HINFO i486 Linux 1.2 donald TXT "DEK" www CNAME ns.linux.bogus richard CNAME ns.linux.bogus ftp A 127.0.0.5 ftp MX 10 mail.linux.bogus ftp MX 20 mail.friend.bogus ftp HINFO P6 Linux 1.3.59 ns A 127.0.0.2 ns MX 10 mail.linux.bogus ns MX 20 mail.friend.bogus ns HINFO Pentium Linux 1.2 ns TXT "RMS" linux.bogus. SOA ns.linux.bogus hostmaster.linux.bogus. (199511301 28800 7200 604800 86400) </verb></tscreen> That's good. Let's check what it says for www alone: <tscreen><verb> > set q=any > www.linux.bogus. Server: localhost Address: 127.0.0.1 www.linux.bogus canonical name = ns.linux.bogus </verb></tscreen> ...In other words, the real name of <tt>www.linux.bogus</tt> is <tt>ns.linux.bogus</tt> <tscreen><verb> linux.bogus nameserver = ns.linux.bogus linux.bogus nameserver = ns.friend.bogus ns.linux.bogus internet address = 127.0.0.2 </verb></tscreen> and ns.linux.bogus has the address 127.0.0.2. Looks good too. <sect1>Winding down Of course, this domain is highly bogus, and so are all the addresses in it, and it is perhaps, unfortunately a bit confusing. For a real example of a real domain see the next section. <sect>A real domain example<label id="real-example"> <bf/Where we list some <em/real/ zone files/ Users have suggested that I include a real example of a working domain as my explanation of what the differences between a working domain and the bogus example was was a bit unclear. One thing about this example: Do <em/not/ enter it into your name servers! Use it only to read for reference. If you want to experiment do that with the bogus example. I use this example with permission from David Bullock of LAND-5. These files were current 24th of September 1996, and might differ from what you find if you query LAND-5's name servers now. Also, keep in mind: delete the leading spaces ;-) <sect1>/etc/named.boot (or /var/named/named.boot) Here we find primary lines for the two reverse zones needed: the 127.0.0 net, as well as LAND-5's 206.6.177 subnet. And a primary line for land-5's forward zone land-5.com. Also note that instead of stuffing the files in a directory called <tt/pz/, as I do in this HOWTO, he puts them in a directory called <tt/zone/. <code> ; Boot file for LAND-5 name server ; directory /var/named ; ; type domain source file or host cache . root.cache primary 0.0.127.in-addr.arpa zone/127.0.0 primary 177.6.206.in-addr.arpa zone/206.6.177 primary land-5.com zone/land-5.com </code> <sect1>/var/named/root.cache Keep in mind that this file is dynamic, and the one listed here is old. You're better off using one produced now, with dig. <code> ; <<>> DiG 2.1 <<>> ;; res options: init recurs defnam dnsrch ;; got answer: ;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 6 ;; flags: qr rd ra; Ques: 1, Ans: 9, Auth: 0, Addit: 9 ;; QUESTIONS: ;; ., type = NS, class = IN ;; ANSWERS: . 518357 NS H.ROOT-SERVERS.NET. . 518357 NS B.ROOT-SERVERS.NET. . 518357 NS C.ROOT-SERVERS.NET. . 518357 NS D.ROOT-SERVERS.NET. . 518357 NS E.ROOT-SERVERS.NET. . 518357 NS I.ROOT-SERVERS.NET. . 518357 NS F.ROOT-SERVERS.NET. . 518357 NS G.ROOT-SERVERS.NET. . 518357 NS A.ROOT-SERVERS.NET. ;; ADDITIONAL RECORDS: H.ROOT-SERVERS.NET. 165593 A 128.63.2.53 B.ROOT-SERVERS.NET. 165593 A 128.9.0.107 C.ROOT-SERVERS.NET. 222766 A 192.33.4.12 D.ROOT-SERVERS.NET. 165593 A 128.8.10.90 E.ROOT-SERVERS.NET. 165593 A 192.203.230.10 I.ROOT-SERVERS.NET. 165593 A 192.36.148.17 F.ROOT-SERVERS.NET. 299616 A 192.5.5.241 G.ROOT-SERVERS.NET. 165593 A 192.112.36.4 A.ROOT-SERVERS.NET. 165593 A 198.41.0.4 ;; Total query time: 250 msec ;; FROM: land-5 to SERVER: default ---- 127.0.0.1 ;; WHEN: Fri Sep 20 10:11:22 1996 ;; MSG SIZE sent: 17 rcvd: 312 </code> <sect1>/var/named/zone/127.0.0 Just the basics, the obligatory SOA record, and a record that maps 127.0.0.1 to <tt/localhost/. Both are required. No more should be in this file. It will probably never need to be updated, unless your nameserver or hostmaster address changes. <code> @ IN SOA land-5.com. root.land-5.com. ( 199609203 ; Serial 28800 ; Refresh 7200 ; Retry 604800 ; Expire 86400) ; Minimum TTL NS land-5.com. 1 PTR localhost. </code> <sect1>/var/named/zone/land-5.com Here we see the mandatory SOA record, the needed NS records. We can see that he has a secondary name server at ns2.psi.net. This is as it should be, <em/always/ have a off site secondary server as backup. We can also see that he as a master host called land-5 which takes care of all the different services, and that he's done it with CNAMEs (a alternative is using A records). As you see from the SOA record, the zone file originates at land-5.com, the contact person is <tt/root@land-5.com/. <tt/hostmaster/ is another oft used address for the contact person. The serial number is in the customary yyyymmdd format with todays serial number appended; this is probably the sixth version of zone file on the 20th of September 1996. Remember that the serial number <em/must/ increase monotonically, here there is only <em/one/ digit for todays serial#, so after 9 edits he has to wait until tomorrow before he can edit the file again. Consider using two digits. <code> @ IN SOA land-5.com. root.land-5.com. ( 199609206 ; serial, todays date + todays serial # 10800 ; refresh, seconds 7200 ; retry, seconds 10800 ; expire, seconds 86400 ) ; minimum, seconds NS land-5.com. NS ns2.psi.net. MX 10 land-5.com. ; Primary Mail Exchanger localhost A 127.0.0.1 router A 206.6.177.1 land-5.com. A 206.6.177.2 ns CNAME land-5.com. ftp CNAME land-5.com. www CNAME land-5.com. mail CNAME land-5.com. news CNAME land-5.com. funn A 206.6.177.3 illusions CNAME funn.land-5.com. @ TXT "LAND-5 Corporation" ; ; Workstations ; ws_177200 A 206.6.177.200 MX 10 land-5.com. ; Primary Mail Host ws_177201 A 206.6.177.201 MX 10 land-5.com. ; Primary Mail Host ws_177202 A 206.6.177.202 MX 10 land-5.com. ; Primary Mail Host ws_177203 A 206.6.177.203 MX 10 land-5.com. ; Primary Mail Host ws_177204 A 206.6.177.204 MX 10 land-5.com. ; Primary Mail Host ws_177205 A 206.6.177.205 MX 10 land-5.com. ; Primary Mail Host ; {Many repetitive definitions deleted - SNIP} ws_177250 A 206.6.177.250 MX 10 land-5.com. ; Primary Mail Host ws_177251 A 206.6.177.251 MX 10 land-5.com. ; Primary Mail Host ws_177252 A 206.6.177.252 MX 10 land-5.com. ; Primary Mail Host ws_177253 A 206.6.177.253 MX 10 land-5.com. ; Primary Mail Host ws_177254 A 206.6.177.254 MX 10 land-5.com. ; Primary Mail Host </code> Another thing to note is that the workstations don't have individual names, but rather a prefix followed by the two last parts of the IP numbers. Using such a convention can simplify maintenance significantly, but can be a bit impersonal, and, in fact, be a source of disgruntlement among your customers. <sect1>/var/named/zone/206.6.177 I'll comment on this file after it. <code> @ IN SOA land-5.com. root.land-5.com. ( 199609206 ; Serial 28800 ; Refresh 7200 ; Retry 604800 ; Expire 86400) ; Minimum TTL NS land-5.com. NS ns2.psi.net. ; ; Servers ; 1 PTR router.land-5.com. 2 PTR land-5.com. 3 PTR funn.land-5.com. ; ; Workstations ; 200 PTR ws_177200.land-5.com. 201 PTR ws_177201.land-5.com. 202 PTR ws_177202.land-5.com. 203 PTR ws_177203.land-5.com. 204 PTR ws_177204.land-5.com. 205 PTR ws_177205.land-5.com. ; {Many repetitive definitions deleted - SNIP} 250 PTR ws_177250.land-5.com. 251 PTR ws_177251.land-5.com. 252 PTR ws_177252.land-5.com. 253 PTR ws_177253.land-5.com. 254 PTR ws_177254.land-5.com. </code> The reverse zone is the bit of the setup that seems to cause the most grief. It is used to find the host name if you have the IP number of a machine. Example: you are an irc server and accept connections from irc clients. However you are a Norwegian irc server and so you only want to accept connections from clients in Norway and other Scandinavian countries. When you get a connection from a client the C library is able to tell you the IP number of the connecting machine because the IP number of the client is contained in all the packets that are passed over the network. Now you can call a function called gethostbyaddr that looks up the name of a host given the IP number. Gethostbyaddr will ask a DNS server, which will then traverse the DNS looking for the machine. Supposing the client connection is from ws_177200.land-5.com. The IP number the C library provides to the irc server is 206.6.177.200. To find out the name of that machine we need to find 200.177.6.206.in-addr.arpa. The DNS server will first find the arpa. servers, then find in-addr.arpa. servers, following the reverse trail through 206, then 6 and at last finding the server for the 177.6.206.in-addr.arpa zone at land-5. From which it will finally get the answer that for 200.177.6.206.in-addr.arpa we have a 'PTR ws_177200.land-5.com' record, meaning that the name that goes with 206.6.177.200 is ws_177200.land-5.com. As with the explanation of how prep.ai.mit.edu is looked up, this is slightly fictitious. Getting back to the irc server example. The irc server only accepts connections from the Scandinavian countries, i.e., *.no, *.se, *.dk, the name ws_177200.land-5.com clearly does not match any of those, and the server will deny the connection. If there was <em/no/ reverse mapping of 206.2.177.200 through the in-addr.arpa zone the server would have been unable to find the name at all and would have to settle to comparing 206.2.177.200 with *.no, *.se and *.dk, none of which will match. Some people will tell you that reverse lookup mappings are only important for servers, or not important at all. Not so: Many ftp, news, irc and even some http (WWW) servers will <em/not/ accept connections from machines that they are not able to find the name of. So reverse mappings for machines are in fact <em/mandatory/. <sect>Maintenance<label id="maint"> <bf/Keeping it working./ There is one maintenance task you have to do on nameds, other than keeping them running. That's keeping the <tt/root.cache/ file updated. The easiest way is using dig, first run dig with no arguments, you will get the <tt/root.cache/ according to your own server. Then ask one of the listed root servers with <tt/dig @rootserver . ns/. You will note that the output looks terribly like a root.cache file except for a couple of extra numbers. Those numbers are harmless. Save it to a file (<tt/dig @e.root-servers.net . ns >root.cache.new/) and replace the old <tt/root.cache/ with it. Remember to restart named after replacing the cache file. Al Longyear sent me this script that can be run automatically to update <tt/root.cache/, install a crontab entry to run it once a month and forget it. The script assumes you have mail working and that the mail-alias `hostmaster' is defined. You must hack it to suit your setup. <code> #!/bin/sh # # Update the nameserver cache information file once per month. # This is run automatically by a cron entry. # ( echo "To: hostmaster <hostmaster>" echo "From: system <root>" echo "Subject: Automatic update of the named.boot file" echo export PATH=/sbin:/usr/sbin:/bin:/usr/bin: cd /var/named dig @rs.internic.net . ns >root.cache.new echo "The named.boot file has been updated to contain the following information:" echo cat root.cache.new chown root.root root.cache.new chmod 444 root.cache.new rm -f root.cache.old mv root.cache root.cache.old mv root.cache.new root.cache ndc restart echo echo "The nameserver has been restarted to ensure that the update is complete." echo "The previous root.cache file is now called /var/named/root.cache.old." ) 2>&1 | /usr/lib/sendmail -t exit 0 </code> Some of you might have picked up that the root.cache file is also available by ftp from Internic. Please <em/don't/ use ftp to update root.cache, the above method is much more friendly to the net. <sect>Bind version 8<label id="bind8"> Bind version 8 is the latest in bind technology. I haven't tried to use it, but David E. Smith (dave@bureau42.ml.org) has. He's written the rest of this section. There's not much to it. Except for using named.conf instead of named.boot, everything is identical. And bind8 comes with a perl script that converts old-style files to new. Example named.boot (old style) for a cache-only name server: <code> directory /var/named cache . root.hint primary 0.0.127.IN-ADDR.ARPA 127.0.0.zone primary localhost localhost.zone </code> On the command line, in the bind8/src/bin/named directory, type: <code> named-bootconf.pl < named.boot > named.conf </code> Which creates named.conf: <code> // generated by named-bootconf.pl options { directory "/var/named"; }; zone "." { type hint; file "root.hint"; }; zone "0.0.127.IN-ADDR.ARPA" { type master; file "127.0.0.zone"; }; zone "localhost" { type master; file "localhost.zone"; }; </code> It works for everything that can go into a named.boot file, although it doesn't add all of the new enhancements and configuration options that bind8 allows. Here's a more complete named.conf that does the same things, but a little more efficiently. <code> // This is a configuration file for named (from BIND 8.1 or later). // It would normally be installed as /etc/named.conf. // The only change made from the `stock' named.conf (aside from this // comment :) is that the directory line was uncommented, since I // already had the zone files in /var/named. options { directory "/var/named"; check-names master warn; /* default. */ datasize 20M; }; zone "localhost" IN { type master; file "localhost.zone"; check-names fail; allow-update { none; }; allow-transfer { any; }; }; zone "0.0.127.in-addr.arpa" IN { type master; file "127.0.0.zone"; check-names fail; allow-update { none; }; allow-transfer { any; }; }; zone "." IN { type hint; file "root.hint"; }; </code> bind8/src/bin/named/test has this, and copies of the zone files, that many people can just drop in and use instantly. The formats for zone files and root.hint (root.cache) files are identical, as are the commands for updating them. <sect>Automatic setup for dialup connections.<label id="auto"> This section explains how I have set things up to automate everything. My way might not suit you at all, but you might get a idea from something I've done. Also, I use ppp for dialup, while many use slip or cslip, so almost everything in your setup can be different from mine. But slip's dip program should be able to do many of the things I do. Normally, when I'm not connected to the net I have a <tt/resolv.conf/ file simply containing the line <tscreen><verb> domain uio.no </verb></tscreen> This ensures I don't have to wait for the host name resolving library to try to connect to a nameserver that can't help me. But when I connect I want to start my named and have a <tt/resolv.conf/ looking like the one described above. I have solved this by keeping two <tt/resolv.conf/ 'template' files named <tt/resolv.conf.local/ and <tt/resolv.conf.connected/. The latter looks like the <tt/resolv.conf/ described before in this document. To automatically connect to the net I run a script called 'ppp-on': <code> #!/bin/sh echo calling... pppd </code> pppd has a file called <tt/options/ that tells it the particulars of how to get connected. Once my ppp connection is up the pppd starts a script called <tt/ip-up/ (this is described in the pppd man page). This is parts of the script: <code> #!/bin/sh interface="$1" device="$2" speed="$3" myip="$4" upip="$5" ... cp -v /etc/resolv.conf.connected /etc/resolv.conf ... /usr/sbin/named </code> I.e. I start my named there. When ppp is disconnected pppd runs a script called <tt/ip-down/: <code> #!/bin/sh cp /etc/resolv.conf.local /etc/resolv.conf read namedpid &etago;var/run/named.pid kill $namedpid </code> So this gets things configured and up when connecting and Dis-configured and down when disconnecting. Some programs, irc and talk come to mind, make a few too many assumptions, and for irc the dcc features and talk to work right you have to fix your hosts file. I insert have this in my <tt/ip-up/ script: <code> cp /etc/hosts.ppp /etc/hosts echo $myip roke >>/etc/hosts </code> <tt/hosts.ppp/ simply contains <code> 127.0.0.1 localhost </code> and the echo thing inserts the ip# i have received for my host name (roke). You should use the name your host knows itself by instead. This can be found with the <tt>hostname</tt> command. It is probably not smart to run named when you are not connected to the net, this is because named will try to send queries to the net and it has a long timeout, and you have to wait for this timeout every time some program tries to resolve a name. If you're using dialup you should start named when connecting and kill it when disconnecting. But please see the <ref id="qanda" name="FAQ"> section for a tip. Some people like to use a forwarders directive on slow connections. If your internet provider has DNS servers at 1.2.3.4 and 1.2.3.5 you can insert the line <code> forwarders 1.2.3.4 1.2.3.5 </code> in the <tt/named.boot/ file.  That will decrease the amount of IP traffic your host originates, any possibly speed things up. This especially important if you're paying pr. byte that goes over the wire. This has the added value of letting you off the one maintenance duty you have as a caching named maintainer; you don't have to update a empty <tt/root.cache/ file. <sect>FAQ<label id="qanda"> In this section I list some of the most frequently asked questions related to DNS and this HOWTO. And the answers :-) Please read this section before mailing me. <enum> <item>How do use DNS from inside a firewall? A couple of hints: `forwarders', `slave', and have a look in the literature list at the end of this HOWTO. <item>How do I make DNS rotate through the available addresses for a service, say www.busy.site to obtain a load balancing effect, or similar? Make several <bf/A/ records for www.busy.site and use bind 4.9.3 or later. Then bind will round-robin the answers. It will <em/not/ work with earlier versions of bind. <item>I want to set up DNS on a (closed) intranet. What do I do? You drop the cache file and just do zone files. That also means you don't have to get new cache files all the time. <item>My system does not have the ndc program. What do I do? Your system then has an old, somewhat obsolete, bind installed. If security is important to you: upgrade bind at once. If not, you can live with it. And instead of running <tt/ndc start/ you run <tt/named/. <tt/ndc reload/ becomes <tt/named.reload/ and <tt/ndc restart/ becomes <tt/named.restart/. All of those programs are most likely in <tt>/usr/sbin</tt>. <item>How do I set up a secondary name server? If the primary server has address 127.0.0.1 you put a line like this in the named.boot file of your secondary: <code> secondary linux.bogus 127.0.0.1 sz/linux.bogus </code> <item>I want bind running when I'm disconnected from the net. I have received this mail from Ian Clark <ic@deakin.edu.au> where he explains his way of doing this: <tscreen><verb> I run named on my 'Masquerading' machine here. I have two root.cache files, one called root.cache.real which contains the real root server names and the other called root.cache.fake which contains... -------------- ; root.cache.fake ; this file contains no information -------------- When I go off line I copy the root.cache.fake file to root.cache and restart named. When I go online I copy root.cache.real to root.cache and restart named. This is done from ip-down & ip-up respectively. The first time I do a query off line on a domain name named doesn't have details for it puts an entry like this in messages.. Jan 28 20:10:11 hazchem named[10147]: No root nameserver for class IN which I can live with. It certainly seems to work for me. I can use the nameserver for local machines while off the 'net without the timeout delay for external domain names and I while on the 'net queries for external domains work normally </verb></tscreen> <item>Where does the caching name server store it's cache? Is there any way I can control the size of the cache? The cache is completely stored in memory, it is <em/not/ written to disk at any time. Every time you kill named the cache is lost. The cache is <em/not/ controllable in any way. named manages it according to some simple rules and that is it. You cannot control the cache or the cache size in any way for any reason. If you want to you can ``fix'' this by hacking named. This is however not recommended. <item>Does named save the cache between restarts? Can I make it save it? No, named does <em/not/ save the cache when it dies. That means that the cache must be built anew each time you kill and restart named. There is <em/no/ way to make named save the cache in a file. If you want you can ``fix'' this by hacking named. This is however not recommended. </enum> <sect>How to become a bigger time DNS admin.<label id="bigger"> <bf>Documentation and tools.</bf> Real Documentation exists. Online and in print. The reading of several of these is required to make the step from small time DNS admin to a big time one. In print the standard book is <em/DNS and BIND/ by C. Liu and P. Albitz from O'Reilly & Associates, Sebastopol, CA, ISBN 0-937175-82-X. I read this, it's excellent. There is also a section in on DNS in TCP/IP Network Administration, by Craig Hunt from O'Reilly..., ISBN 0-937175-82-X. Another must for Good DNS administration (or good anything for that matter) is <em/Zen and the Art of Motorcycle Maintenance/ by Robert M. Prisig :-) Available as ISBN 0688052304 and others. Online you will find stuff on <url url="http://www.dns.net/dnsrd/">, <url url="http://www.isc.org/bind.html">; A FAQ, a reference manual (BOG; Bind Operations Guide) as well as papers and protocol definitions and DNS hacks (these, and most, if not all, of the rfcs mentioned below, are also contained in the bind distribution). I have not read most of these, but then I'm not a big-time DNS admin either. Arnt Gulbrandsen on the other hand has read BOG and he's ecstatic about it :-). The newsgroup <htmlurl url="news:comp.protocols.tcp-ip.domains" name="comp.protocols.tcp-ip.domains"> is about DNS. In addition there are a number of RFCs about DNS, the most important are probably these: <descrip> <tag/RFC 2052/ A. Gulbrandsen, P. Vixie, <em/A DNS RR for specifying the location of services (DNS SRV)/, October 1996 <tag/RFC 1918/ Y. Rekhter, R. Moskowitz, D. Karrenberg, G. de Groot, E. Lear, <em/Address Allocation for Private Internets/, 02/29/1996. <tag/RFC 1912/ D. Barr, <em/Common DNS Operational and Configuration Errors/, 02/28/1996. <tag/RFC 1713/ A. Romao, <em/Tools for DNS debugging/, 11/03/1994. <tag/RFC 1712/ C. Farrell, M. Schulze, S. Pleitner, D. Baldoni, <em/DNS Encoding of Geographical Location/, 11/01/1994. <tag/RFC 1183/ R. Ullmann, P. Mockapetris, L. Mamakos, C. Everhart, <em/New DNS RR Definitions/, 10/08/1990. <tag/RFC 1035/ P. Mockapetris, <em/Domain names - implementation and specification/, 11/01/1987. <tag/RFC 1034/ P. Mockapetris, <em/Domain names - concepts and facilities/, 11/01/1987. <tag/RFC 1033/ M. Lottor, <em/Domain administrators operations guide/, 11/01/1987. <tag/RFC 1032/ M. Stahl, <em/Domain administrators guide/, 11/01/1987. <tag/RFC 974/ C. Partridge, <em/Mail routing and the domain system/, 01/01/1986. </descrip> </article>